Estimates of genotoxic effects of mutagens at low and protracted doses are often based on linear extrapolation of data obtained at relatively high doses. To test the validity of such an approach, a comparison was made between the mutagenicity of N-methyl-N-nitrosourea (MNU) in T-lymphocytes of the rat following two treatment protocols, i.e. sub-chronic exposure to a low dose (15-45 repeated exposures to 1mg/kg of MNU) or acute exposure to a single high dose (15, 30 or 45 mg/kg of MNU). Mutation induction appeared dramatically lower following sub-chronic treatment compared to treatment with a single high exposure. Furthermore, DNA sequence analysis of the coding region of the hprt gene in MNU-induced mutants showed that acute high dose treatment causes mainly GC-->AT base pair changes, whereas sub-chronic treatment results in a significant contribution of AT base pair changes to mutation induction. We hypothesize that O(6)-methylguanine-DNA methyltransferase is saturated after acute treatments, while after sub-chronic treatment most O(6)-methylguanine is efficiently repaired. These data suggest (i) that risk estimations at low and protracted doses of MNU on the basis of linear extrapolation of effects measured at high dose are too high and (ii) that the protective effects of DNA repair processes are relatively strong at low sub-chronic exposure.